Polymer 48 (2007) 2988e3002

www.elsevier.com/locate/polymer

Modeling of vapor sorption in glassy polymers using a new dual mode

sorption model based on multilayer sorption theory
Haidong Feng*
School of Mechanical and Power Engineering, 102, Shanghai Jiao Tong University, 1954 Huashan Road, Shanghai 200030, PR China
Received 6 January 2006; received in revised form 24 October 2006; accepted 31 October 2006
Available online 5 April 2007

Abstract

Conventional dual mode sorption (CDMS) model is one of the most effective models in describing vapor sorption isotherms with a concave
towards the activity axis in glassy polymers, while engaged species induced clustering (ENSIC) model has been approved to be highly successful
in modeling vapor sorption isotherms in polymers with a convex to the activity axis (BET type III) over a wide range. However, neither of them
is effective to describe other types of vapor sorption isotherms, especially sigmoidal isotherms. The GuggenheimeAndersonede Boer (GAB)
model fits extremely well with sigmoidal isotherms such as some vapor especially water vapor sorption data in food and related natural mate-
rials. However, one assumption of the GAB model for vapor sorption in glassy polymers is inconsistent with the fact that there are two species of
sorption sites as the CDMS model assumes. Based on multilayer sorption theory on which the GuggenheimeAndersonede Boer (GAB) model is
based, a new dual mode sorption (DMS) model for vapor sorption in the glassy polymers is deduced. The mathematical meanings and the phys-
icochemical significances of the parameters in the new model are analyzed. The new model has been verified experimentally by some special
cases. Comparisons of the new DMS model with the CDMS and the ENSIC models prove that only the new model fits extremely well with all
types of vapor sorption isotherms in the glassy polymers.
Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Glassy polymer; Vapor sorption; GAB model

1. Introduction convex at high activities [4e10], as illustrated in Fig. 1(b).

There are three main types of vapor sorption isotherms in
glassy polymers, as shown in Fig. 1(a)e(c), where c and
a are, respectively, penetrant concentration and activity in the
polymers.
Type 1 is concave to activity axis at low activities and
almost linear at higher activities [1e3], as shown in Fig. 1(a).
Most of the vapor sorption isotherms in glassy polymers are
Type 1, such as water vapor sorption isotherms in two poly-
imide copolymers, PMDAe50DDS/50ODA and BPDAe
50DDS/50ODA at 30  C [2], as illustrated in Fig. 13.
Type 2 regularly referred to as BET type II, is a sigmoidal
isotherm, which is concave to the abscissa at low activities and
* Tel.: þ86 21 62934217; fax: þ86 21 64078095.
E-mail address: fhdfhd@sjtu.edu.cn
There is an inflection point in the isotherm. Many vapor
sorption isotherms in glassy polymers are this type, such as
methanol vapor sorption in cellulose acetate [10], as shown
in Fig. 16.
Type 3 is always convex to the activity axis [6,10], as
shown in Fig. 1(c). Only a few of vapor sorption isotherms
in glassy polymers belong to this type, such as ethanol vapor
sorption in cellulose acetate [10], as illustrated in Fig. 18.
Two special isotherms, Type 4 and Type 5, both of which
are Type 3 actually, should be specially mentioned. Type 4
looks like Type 1, because the downward curvature at rela-
tively high activities is too inconspicuous to be discerned,
such as water and ethanol vapor sorption in poly(vinylchlor-
ide) (PVC) at 40  C [11], as shown in Fig. 14. Type 5 looks
like Type 3, because the upward curvature at low activities
is too inconspicuous to be discerned, such as water vapor sorp-
tion in cellulose acetate at 25  C [10], as displayed in Fig. 17.

0032-3861/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.polymer.2006.10.050
 H. Feng / Polymer 48 (2007) 2988e3002 2989

List of symbols s2 average number of molecules per sorbed site in

a penetrant activity in a polymer microvoid region of a polymer
A temperature-dependent constant in GAB model s1 average values of average number of sorbed
A0 pre-exponential factor of a temperature-depen- molecules per site in matrix region of a polymer
dent constant in GAB model over the entire range of activity
A0 temperature-dependent constant in new DMS s2 average values of average number of sorbed
model molecules per site in microvoids of a polymer
b microvoid affinity constant based on pressure over the entire range of activity
b0 microvoid affinity constant based on activity T temperature
c penetrant concentration in a glassy polymer Tg glass transition temperature
c1 penetrant concentration in matrix region of a yi experiment value
glassy polymer by new DMS model byi predicted value
c2 penetrant concentration in microvoids of a glassy yi mean value
polymer by new DMS model
cD penetrant concentration in matrix region of List of abbreviations
a glassy polymer by CDMS model AP pyromellitic anhydride
cH penetrant concentration in microvoids of a glassy BET Brunauer, Emmet and Teller
polymer by CDMS model BPDA biphenyl tetracarboxylic dianhydride
CH0 Langmuir saturation constant CDMS conventional dual mode sorption
Cp monolayer sorption capacity DMS dual mode sorption
Cp1 monolayer sorption capacity in matrix region of ENSIC Engaged species induced clustering
a polymer GAB GuggenheimeAndersonede Boer
Cp2 monolayer sorption capacity in microvoid region MDI macrodiisocyanate
of a polymer NTDAeoBAPBDS a copolymer by copolymerization of
Cp0 weighted mean value of sorption capacity of 1,4,5,8-naphthalenetetracarboxylic dianhydride
a polymer to sorbate molecules (NTDA) and 2,20 -bis(4-aminophenoxy)biphenyl-
Cp weighted mean value of polymer sorption capac- 5,50 -disulfonic acid (oBAPBDS)
ity to sorbate molecules based on s1 and s2 NTDAeBAPBDS a copolymer by copolymerization of
H heat given off when a molecule enters sorbed 1,4,5,8-naphthalenetetracarboxylic dianhydride
sites of a polymer (NTDA) and 4,40 -bis(4-aminophenoxy)biphenyl-
HL heat of condensation of a pure vapor 3,30 -disulfonic acid
Hm heat of sorption of monolayer of a vapor NTDAeoBAPBDS/mBAPPS(2/1) a copolymer by copoly-
Hn heat of sorption of multimolecular layers of a vapor merization of 1,4,5,8-naphthalenetetracarboxylic
k temperature-dependent constant in GAB model dianhydride (NTDA) and 2,20 -bis(4-aminophen-
k0 temperature-dependent constant in new DMS oxy)biphenyl-5,50 -disulfonic acid (oBAPBDS)/bis
model [4-(3-aminophenoxy)-phenyl] sulfone (mBAPPS)
k0 pre-exponential factor of a temperature-depen- ODA oxydianiline
dent constant in GAB model PCL poly(caprolactonediol)
kD Henry’s law dissolution constant based on pressure PDMS poly(dimethylsiloxane)
kD0 Henry’s law dissolution constant based on activity PET poly(ethylene terephthalate)
kp affinity of a penetrant molecule towards a poly- PMDA pyromellitic dianhydride
mer site PPG poly(propyleneglycol)
ks affinity of a penetrant molecule towards a like PPO poly(propylene oxide)
molecule PTMG poly(tetramethyleneglycol)
n n-layer sorption PUI poly(urethaneimide)
np polymer segment number PVC poly(vinylchloride)
ns sorbed solvent molecule number SSE sum of squares due to error
p pressure SSR sum of squares of regression
p0 vapor saturation pressure SST total sum of squares
R gas constant TFE/BDD65 a glassy copolymer containing 65 mol%
R2 square of the correlation between the response 2,2-bis (trifluoromethyl)-4,5-difluoro-1,3-diox-
values and the predicted response values ole (BDD) and 35 mol% tetrafluoroethylene
s average number of molecules per sorbed site TFE/BDD87 a glassy copolymer containing 87 mol%
s1 average number of molecules per sorbed site in 2,2-bis (trifluoromethyl)-4,5-difluoro-1,3-diox-
matrix region of a polymer ole (BDD)
2990 H. Feng / Polymer 48 (2007) 2988e3002

Fig. 1. Vapor sorption isotherms in glassy polymers.

Conventional dual mode sorption (CDMS) model is known
to be quite effective in describing Type 1 sorption isotherms of
gas or vapor in the glassy polymers [12e18]. It assumes that
penetrant molecules are partitioned into two distinct popula-
tions, which are in dynamic equilibrium with each other: (i)
penetrant molecules sorbed by a dissolution mechanism in
the equilibrium dense polymer matrix (Henry’s law popula-
tiondjust like gas sorption in rubbery polymers) and (ii)
penetrant molecules filling the non-equilibrium unrelaxed,
molecular-scale gaps (microvoids) frozen into the glassy state
(Langmuir population), as shown in Fig. 2. However, this
model cannot describe Types 2 and 3 sorption isotherms
effectively.
Engaged species induced clustering (ENSIC) model has
been approved to be highly successful in modeling Type 3
vapor sorption isotherms and is better than the conventional
FloryeHuggins and related models [19,20]. However, it is
unsatisfactory in modeling Type 2 sorption isotherms.
Despite having appeared in an article as early as 1946, the
GuggenheimeAndersonede Boer (GAB) model has agreed
extremely well with Type 2 isotherms such as vapor especially
water vapor sorption data in food and related natural materials
quite recently [21e24]. However, the assumption of the GAB
model that all the sorption sites are equivalent is inconsistent
with the glassy polymers which are considered to have two
species of sorption sites as the CDMS model assumes.
The objective of present work is to develop a model which
can describe all the types of isotherms of vapor sorption in the
glassy polymers effectively and reasonably. To this end, based
on multilayer sorption theory on which the GAB equation is
based, a new dual mode sorption (DMS) model for vapor sorp-
tion in the glassy polymers is deduced. One assumption has
been made in this model that the vapor sorption in the poly-
mers is composed of two parts, one of which occurs in the
equilibrium matrix region of the glassy polymers which is
a downward curve similar to that in rubbery polymers, the
other happens in the non-equilibrium microvoids which is an
upward curve. The mathematical meanings and the physico-
chemical significances of the parameters in the new model
are analyzed. The new model is proved to agree extremely
well with all types of vapor sorption isotherms known in the
glassy polymers in the present work.
2. Theoretical backgrounds
2.1. CDMS and ENSIC models
According to the CDMS model, the penetrant concentration
in the glassy polymers is expressed analytically as a sum of
two contributions to penetrant sorption [12e18]:
c ¼ cD þ c H ð1Þ

cD and cH are given as

cD ¼ k D p ð2Þ

C0H bp
cH ¼ ð3Þ
1 þ bp
Then, c becomes
C0H bp
c ¼ kD p þ ð4Þ
1 þ bp

where cD and cH are the penetrant concentrations by Henry’s

law mode and Langmuir mode, respectively; kD is the Henry’s
law dissolution constant; b is the microvoid affinity constant;
C0H is the Langmuir saturation constant.
For vapor sorption, the original CDMS equation (Eq. (4))
also can be rewritten as

p C0H bp0 pp0 C0H b0 a

c ¼ k D p0 þ p ¼ kD0 a þ ð5Þ
Fig. 2. Schematic of the surface of a glassy polymer.
p0 1 þ bp0 p0 1 þ b0 a
 H. Feng / Polymer 48 (2007) 2988e3002 2991

where kD0 ¼ kD p0 , which is actually penetrant activity coeffi-
cient at finite dilution in the matrix region of the glassy poly-
mer; b0 ¼ bp0 is the microvoid affinity constant in the form of
activity.
Then cD and cH are expressed as
Therefore, both the CDMS and ENSIC models are ineffec-
tive in modeling sigmoidal isotherms in that there are no
inflection points in them.
2.2. New dual mode sorption model

cD ¼ kD0 a ð6Þ 2.2.1. Statistical multilayer sorption theory GAB model

based on
C0H b0 a It is well known that the original GAB model can be de-
cH ¼ ð7Þ rived by kinetic as well as by statistical approaches. The pres-
1 þ b0 a
ent work utilizes the statistical one. Let the sorption system
The second-order derivative of Eq. (5) with respect to a is consist of X sorbate molecules bound to an array of N sorption
then given as sites. The first shell of vapor covers the glassy polymer surface
d2 c 2C0H b20 and is tightly bound in a monolayer. Subsequent layers of
¼  <0 ð8Þ
da2 ð1 þ b0 aÞ3 vapor have weaker interactions with the polymer surface. The
interactions between the molecules in these layers and the
This illustrates that the CDMS equation is an upward curve polymer range in energy level s somewhere between those
and has no inflection point. Hence, the CDMS model cannot of the first layer molecules and the bulk liquid.
be utilized to describe Type 2 and Type 3 vapor sorption For the derivation of the GAB equation, some assumptions
isotherms effectively. are made. The sites are equivalent, distinguishable and inde-
The ENSIC model is, in fact, derived from a rather simple pendent, and their spatial arrangement is immaterial. On
mechanistic approach assuming two elementary sorption each site, there are s number of molecules ð0  s  NÞ which
processes for a molecule in the vapor phase [19,20]: can be sorbed in a vertical pile without interactions between
molecules sorbed to different sites, i.e. there exists no horizon-
(1) condensation on a polymer site with a probability related tal interactions. The positions of the molecules in the sorption
to the affinity, kp, of a penetrant molecule towards a poly- pile are distinguishable and the thermodynamic state (chemi-
mer site; and cal potential) of the sorbed molecules is regulated by an outer
(2) condensation on a previously sorbed penetrant molecule gas phase at a given constant temperature T and variable
related to the affinity, ks, of a penetrant molecule towards pressure p. Under all these assumptions, the grand partition
a like molecule. function, X, of the system is related to the partition function,
Q, of each site as
The relationship between the increase of sorbed solvent
molecule number dns and the pressure increase in the gaseous N
phase dp can be expressed as XðN; T; mÞ ¼ QðT; mÞ ð12Þ

  and Q is given as
dns ¼ kp np þ ks ns dp ð9Þ

where np and ns are the polymer segment number and the Q ¼ 1 þ j1 l þ j1 j2 l2 þ j1 j2 j3 l3 þ / þ j1 j2 .jn ln ð13Þ
solvent number in the polymer matrix, respectively.
Integration of Eq. (9) gives the function of the ENSIC where m is the chemical potential of the molecules of
model [19] the sorbed vapor; l is the corresponding absolute activity,
  and l ¼ expðm=RTÞ; ji ð1  i  nÞ is the partition function
kp     
c¼ exp ks  kp a  1 ð10Þ of a molecule in layer i of the sorption pile. Then the average
ks  kp number, X, of molecules on the sorbed sites in the polymer is
If there is no preferential affinity between kp and ks, i.e. expressed as
kp ¼ ks, integration of Eq. (9) gives linear relationship between
     
c and a, and Henry’s dissolution occurs. vln X vln X vln Q
The second-order derivative of Eq. (10) with respect to a is X ¼ kT ¼l ¼ Nl ð14Þ
vm N;T vl N;T vl T
then given as
The average number, s, of the molecules per site is given as
2
dc      
¼ kp ks  kp exp ks  kp a ð11Þ vlnQ
da2 s ¼ X=N ¼ l
vl T
d2 c j1 l þ 2j1 j2 l2 þ 3j1 j2 j3 l3 þ / þ n j1 j2 .jn ln
da2 s0 unless the relationship between a and c is linear,
¼ ð15Þ
i.e. kp ¼ ks. This indicates that the ENSIC
equation also has no 1 þ j1 l þ j1 j2 l2 þ j1 j2 j3 l3 þ / þ j1 j2 .jn ln
inflection point. Hence, the ENSIC model cannot be utilized
to describe Type 2 vapor sorption isotherms effectively. ji can be calculated as
2992 H. Feng / Polymer 48 (2007) 2988e3002

j1 ¼ Aj ¼ Akj0 ð16Þ number of sorption sites of the polymer and the properties
of sorbed molecules.
j2 ¼ j3 ¼ / ¼ jn ¼ kj0 ð17Þ If n is large enough, Eq. (24) simplified as

where j0 is the partition function of the sorbed molecules in the Cp Aka

c¼ ð25Þ
pure liquid state. ð1  kaÞð1  ka þ AkaÞ
The parameter k is defined as the ratio of the partition func-
tion of molecules sorbed in the multilayer to that of molecules This is the famous GAB equation.
in bulk liquid. It is called correction factor, since it corrects the In the kinetic approach, parameter A measures the difference
properties of the multilayer molecules relative to the bulk between molecules in the monolayer and in the multilayer, while
liquid or indicates the difference between the liquid molecules parameter k measures the difference between molecules in
and multilayer molecules. Parameter A is defined as the ratio the multilayer and in the bulk liquid. k and A are calculated as
of the partition function of the first molecule sorbed on  
a site to that of molecules sorbed beyond the first molecule HL  Hn
k ¼ k0 exp ð26Þ
in the multilayer. It is a measure of the strength of binding RT
of sorbed molecules to the primary binding sites. Larger the  
A, stronger the sorbed molecules bound in the mono- Hm  Hn
A ¼ A0 exp ð27Þ
layer and larger the difference between the monolayer mole- RT
cules and multilayer molecules. n is the number of sorbed
layers. where k0 and A0 are pre-exponential factors; HL is the heat of
Then, Eq. (15) becomes condensation of pure vapor; Hm and Hn are the heat of sorption
 of monolayer and multimolecular layers (n layers) of the
2 n1 
cfj0 l 1 þ 2jn l þ 3ð jn lÞ þ/ þ nð jn lÞ vapor, respectively.
s¼  
1 þ cfj0 l 1 þ jn l þ ð jn lÞ2 þ/ þ ð jn lÞn1 Since the multilayer molecules are less firmly bound than
Pn i1 the bulk liquid molecules, k will be less than unity.
cfj0 l i¼0 ið jn lÞ There are three special cases for the multilayer sorption
¼ Pn i1 (18)
1 þ cfj0 l i¼1 ð jn lÞ isotherms.
Using the summation formulae of finite geometrical If n ¼ 1, i.e. it is a monolayer sorption, Eq. (25) is simpli-
progressions, fied as
 nþ1  Cp Aka
X
n
i1
n
1  ðn þ 1Þð jn lÞ þnð jn lÞ c¼ ð28Þ
ið jn lÞ ¼ 2 ð19Þ 1 þ Aka
i¼0 ð1  jn lÞ
This equation is just like Langmuir adsorption equation.
and If the first layer molecules have the same partition functions
as those of the other layers or the first layer molecules have the
Xn
1  ð jn lÞ
n
ð jn lÞi1 ¼ ð20Þ same heat of sorption as those in other layers, i.e. A ¼ 1, and
i¼1
1  jn l n is large enough, Eq. (24) or (25) is simplified as
Cp ka
and identifying c¼ ð29Þ
1  ka
j0 l ¼ a ¼ p=p0 ð21Þ If the layers other than the first layer have the same parti-
tion functions as those of the pure liquid, i.e. k ¼ 1, Eq. (25)
jn l ¼ ka ð22Þ becomes
Eq. (18) becomes Cp Aa
c¼ ð30Þ
 nþ1  ð1  aÞð1  a þ AaÞ
Aka 1  ðn þ 1ÞðkaÞn þnðkaÞ
s¼ n ð23Þ
ð1  kaÞð1  ka þ Akað1  ðkaÞ ÞÞ This is the famous Brunauer, Emmett and Teller (BET)
The sorbate concentration in the polymer is then obtained equation.
as
  2.2.2. New dual mode sorption model
Cp Aka 1  ðn þ 1ÞðkaÞn þnðkaÞnþ1 Four assumptions are made for the new dual mode sorption
c ¼ Cp s ¼ n ð24Þ
ð1  kaÞð1  ka þ Akað1  ðkaÞ ÞÞ model:

where Cp is the monolayer sorption capacity, which is the va- (1) The sorption sites in a glassy polymer material can be di-
por content in the polymer corresponding to each sorption site vided into two different species; one is located in the ma-
occupied by one vapor molecule. It mainly depends on the trix region of the polymer and the other in the microvoids.
 H. Feng / Polymer 48 (2007) 2988e3002
(2) All the molecules in the matrix region of the polymer have
the same partition functions.
(3) It is GAB sorption in the microvoid region of the polymer.
(4) The molecules in the layers other than the first in the mi-
crovoid region have the same partition function as those in
the matrix region.
According to the second assumption, sorption in the matrix
region as follows:
Cp1 k1 a
c1 ¼ Cp1 s1 ¼
1  k1 a
Cp1 is the monolayer sorption capacity in the matrix region of
the polymer, which is interpreted as the vapor content corre-
sponding to each sorption site occupied by one vapor molecule
in this region. k1 measures the difference between molecules in
the multilayer and in the bulk liquid in the matrix region of the
polymer.
This assumption is suitable for vapor molecules diffusing
through the polymer both individually and in clusters.
Fig. 3 illustrates vapor molecules diffusing through the
polymer individually. n vapor molecules sequentially sorb on
the surface of the polymer and then diffuse through the poly-
mer, as shown in Fig. 3(a). Then all the vapor molecules will
have the same partition functions or have the same heat of
sorption in this case. It is equivalent to that the n molecules
which are sorbed in a vertical pile have the same partition
function as that of the first molecule in the matrix region of
the polymer, as shown in Fig. 3(b).
Fig. 4 illustrates sorption in the matrix region of a polymer
for vapor molecules which diffuse through the polymer in
clusters. Provided that three molecules are in a cluster, they
sorb on the surface of the polymer, as shown in Fig. 4(a).
Fig. 3. Schematic of sorption in the matrix region of a polymer for vapor molecules which diffuse through the polymer individually.
Fig. 4. Schematic of sorption in the matrix region of a polymer for vapor molecules which diffuse through the polymer in clusters.
2993
Then each of the molecules will have the same partition func-
tion. This is equivalent to that the three molecules which are
sorbed in a vertical pile have the same partition functions as
that of the first molecule in the matrix region of the polymer,
as shown in Fig. 4(b).
Of course, if m clusters sequentially sorb on the surface of
the polymer and then diffuse through the polymer, the assump-
Cp1 k1 a
tion holds and the sorption follows c1 ¼ Cp1 s1 ¼ 1k 1a
.
In terms of the third assumption, sorption in the microvoid
region as follows:
ð31Þ
Cp2 Ak2 a
c2 ¼
ð1  k2 aÞð1  k2 a þ Ak2 aÞ
 
1 ðA  1Þk2 a
¼ Cp2 þ ð32Þ
1  k2 a 1 þ ðA  1Þk2 a
Cp2 is the monolayer sorption capacity in the microvoid region
of the polymer. k2 measures the difference between molecules
in the multilayer and in the bulk liquid in the microvoid region
of the polymer. A measures the difference between molecules
in the monolayer and in the multilayer in the microvoid region.
The sorption in this region is essentially divided into two
kinds: one is just like the sorption in the matrix region, which
follows c21 ¼ Cp2 1k1 2 a; the other is totally dependent on the
ðA1Þk2 a
microvoids, which follows c22 ¼ Cp2 1þðA1Þk 2a
.
On one hand, if A ¼ 1, which means the first layer mole-
cules have the same partition functions as those of the other
layers or the first layer molecules have the same heat of sorp-
tion as those in other layers in the microvoid region, only one
mode of sorption occurs on the surface of the polymer. In ad-
dition, if A ¼ 1, c2e2 ¼ 0, which means the sorption which is
totally dependent on the microvoids disappears. The polymer
should be in its rubbery state. On the other hand, if a polymer
2994 H. Feng / Polymer 48 (2007) 2988e3002

is in its rubbery state, microvoids in the polymer disappear. k0 a

Then, c2e2 ¼ 0, and A ¼ 1 accordingly. c1 ¼ Cp s01 ¼ Cp ð42Þ
1  k0 a
According to the third assumption, the following equation
is obtained. ðA0  1Þk0 a
c2 ¼ Cp s02 ¼ Cp ð43Þ
k1 ¼ k2 ¼ k ð33Þ 1 þ ðA0  1Þk 0 a
In terms of the first assumption, the sorbate concentration
where s01 and s02 are the average values of the average number
in the polymer is given as
of sorbed molecules per site in the matrix region and micro-
c ¼ c1 þ c2 ¼ Cp1 s1 þ Cp2 s2 voids of the polymer based on Cp , respectively; k0 is the ratio
  of the partition function of molecules sorbed in the multilayer
ka ka ðA  1Þka
¼ Cp1 þ Cp2 þ ð34Þ to that of molecules in bulk liquid based on Cp ; A0 is the ratio
1  ka 1  ka 1 þ ðA  1Þka of the partition function of the first molecule sorbed on a site
to that of molecules sorbed beyond the first molecule in the
It is a four-parameter equation.
multilayer based on Cp .
Since the surface area of the matrix region is much greater
The sorbate concentration in the polymer is given as
than that of microvoid region, the following inequality holds.
ka ka k0 a ðA0  1Þk0 a
Cp1  Cp2 ð35Þ c ¼ Cp þ C p ð44Þ
1  ka 1  ka 1  k0 a 1 þ ðA0  1Þk 0 a
Then Eq. (34) becomes
ka ðA  1Þka This is the new DMS equation which has three parameters.
c ¼ Cp1 þ Cp2 ð36Þ Therefore, in terms of Eq. (44), the new DMS model for vapor
1  ka 1 þ ðA  1Þka
sorption in glassy polymers can be explained as follows. The
Since both Cp1 and Cp2 are mainly dependent on the sorption is contributed to two modes, as shown in Fig. 5,
properties of the polymer and the vapor, let one of which occurs in the matrix region of the glassy poly-
k0 a
mers and follows c1 ¼ Cp 1k 0 a which is a downward curve sim-
c ¼ C0p ðs1 þ s2 Þ ð37Þ
ilar to that in rubbery polymers; 0 the 0 other happens in the
ðA 1Þk a
Then, microvoids and follows c2 ¼ Cp 1þðA 0 1Þk 0 a which is an upward

s1 s2 curve. The total sorption amount of vapor molecules in the

C0p ¼ Cp1 þ Cp2 ð38Þ glassy polymers is equal to the sum of the two species.
s1 þ s2 s1 þ s2
As stated before, Cp is the weighted mean value of the
Cp0 is the weighted mean value of the sorption capacity of the sorption capacity of a polymer to a vapor. It mainly depends
polymer to the sorbate. on the structure and the state of the polymer. Besides, the size
Since s1 and s2 are functions of a, Cp0 is not a constant over of the vapor molecule also has some influence on the sorption
the entire range of activity. capacity. For example, there are three sorption sites for larger
Let vapor A molecules, and eight for smaller vapor B molecules,
R1 Z 1 as illustrated in Fig. 6.
s1 da According to both the statistical and kinetic approaches, k 0
s1 ¼ 0 ¼ s1 da ð39Þ
10 0 is the difference between the interaction of vapor molecules
R1 (shown in Fig. 7(a)) and that of the vapor molecule and a poly-
Z 1
0
s1 da mer molecule segment (shown in Fig. 7(b)). It is essentially an
s2 ¼ ¼ s2 da ð40Þ indication of the interaction between the vapor molecule and
10 0
the polymer molecule segment. Larger k0 indicates stronger
where s1 and s2 are the average values of the average number of interaction between the vapor and the polymer.
sorbed molecules per site in the matrix region and microvoids
of the polymer over the entire range of activity, respectively.
Let
s1 s2
Cp ¼ Cp1 þ Cp2 ð41Þ
s1 þ s2 s1 þ s2

Cp is the weighted mean value of the sorption capacity of the

polymer to the sorbate based on s1 and s2 . It is a constant at
a given temperature. If s2 ¼ s1 , the sorption site in the
microvoid region is equivalent to that in the matrix region.
Otherwise, they are different. For example, if s2 ¼ 2s1 ,
one sorption site in the microvoid region is equivalent to two
sites in the matrix region.
Then, c1 and c2 become Fig. 5. Schematic representation of the new DMS model.
 H. Feng / Polymer 48 (2007) 2988e3002 2995

Fig. 6. Influences of the sizes of the vapor molecules on the sorption capacity.

The mathematical meaning of k0 is to demonstrate the mag-

nitude and the curvature of c1, as illustrated in Fig. 8. Larger
the value of k0 , stronger the interaction between polymer and
vapor, greater the departure from linearity, and more the
sorbed vapor molecules in the polymer, and vice versa. For ex-
ample, water vapor/hydrophilic polymer systems have higher Fig. 8. Schematic of influences of parameter k0 on c1.
values of k0 , while water vapor/hydrophobic polymer systems
have lower values of k0 .
As analyzed before, A0 is the difference between the inter- b0 s0 for Eq. (7), while A0 ¼ 1, Cp s0 for Eq. (43) or (44).
action of a microvoid and the first molecule sorbed on it and Therefore, Eq. (43) is not a strictly Langmuir adsorption.
those of the microvoid and the molecules sorbed beyond the Eq. (44) has a similar form to the original GAB equation.
first molecule in the multilayer. In terms of the assumptions However, the GAB equation is used to describe multilayer
of the new model, the interactions of the microvoid and the sorption while Eq. (44) is utilized to describe the vapor/glassy
molecules sorbed beyond the first molecule are equal to those polymer systems as dual mode sorption. The GAB equation
of the molecules and the polymer molecule segments, as assumes that all the sorption sites are equivalent while there
shown in Fig. 9(a). Therefore, A0 is actually a measure of are two different species of sorption sites in the new DMS
the interaction of the vapor molecule and the microvoid model.
(shown in Fig. 9(b)). The relationship between the CDMS model and the new
Moreover, according to Eq. (43), ðA0  1Þk 0 measures the model is analyzed as follows.
affinity between the vapor and the microvoid of the polymer. If k0 1, Eq. (42) becomes
Since the interaction between the vapor and a polymer in the
c1 ¼ Cp k0 a ð45Þ
matrix region and that in the microvoid region share the
same k0 , the microvoid properties should be demonstrated Eq. (44) then changes to
only by A0.
Eq. (43) has a similar form to the Langmuir sorption iso-
therm Eq. (7). However, for rubbery polymers, C0H ¼ 0 and

Fig. 9. Schematic of the interaction between the vapor molecule and a polymer
Fig. 7. Schematic of the interaction between vapor molecules (a) and between molecule segment (a) and that between the vapor molecule and a microvoid of
of the vapor molecule and a polymer molecule segment (b). the polymer (b).
2996 H. Feng / Polymer 48 (2007) 2988e3002

ðA0  1Þk0 a and SSR is defined as

c ¼ Cp k0 a þ Cp ð46Þ
1 þ ðA0  1Þk 0 a
X
n
 2
Let k1 ¼ Cpk0 , and b0 ¼ (A0 1)k0 , Eq. (46) becomes SSR ¼ byi  yi ð52Þ
i¼1
b0 a
c ¼ k1 a þ Cp ð47Þ
1 þ b0 a SST is
It has the same form as the CDMS model (Eq. (4)). There-
fore, the CDMS model can be considered to be a special case X
n
2
SST ¼ ðyi  yi Þ ð53Þ
of the new DMS model. i¼1
If A0 ¼ 1, c ¼ c1 in terms of Eq. (44), indicating that the
microvoids in the polymers disappear and the polymers are where yi is the mean value.
in their rubbery states. Therefore, a rubbery polymer can
be also considered to be a special case of the new model. 3. Results and discussion
The second-order derivative of Eq. (42) with respect to a is
then given as 3.1. Verification of the new DMS model
2 02
d c1 2Cp k
2
¼ 3 >0 ð48Þ It is extremely difficult to verify experimentally the dual
da ð1  k0 aÞ mode sorption models directly, whether it is the conventional
Hence, Eq. (42) is strictly convex, that is, vapor isotherm in or the new, in that it is almost impossible to measure vapor
the equilibrium matrix region of the polymer is strictly convex concentrations in the polymer matrix and in the microvoid re-
over the entire region of activities. gion separately. However, the experiment data can be obtained
The second-order derivative of Eq. (43) with respect to a is for a special case, A0 ¼ 1 which indicates that the microvoids
then given as in the polymer disappear and the polymer should be in its rub-
bery state. In addition, if the assumptions of the new DMS
d2 c 2 2Cp b02 model for vapor sorption in polymers are valid, c1 ¼ Cp 1k k0 a
¼  3 0 ð49Þ 0a
da2 ð1 þ b0 aÞ can fit the experiment data well. On the other hand, if a poly-
Eq. (43) is concave, meaning that the vapor isotherm in the mer is in its rubbery state, A0 should be equal to 1 when the
non-equilibrium microvoids of the polymer is concave over vapor sorption in the polymer is modeled by the new DMS
the entire region of activities. equation. Moreover, A0 s0 for vapor sorption in a polymer
Therefore, Eqs. (42) and (43) have opposite convexities. with Type 3 isotherm indicates that the polymer is glassy.
The new DMS model can be utilized to describe all the
three types of vapor sorption isotherms, shown in Fig. 1, in 3.1.1. Example 1
the glassy polymers. Fig. 10 presents isotherms of water vapor sorption in nylon
6,6 at 25  C (B), 35  C (6), and 45  C (7). The solid lines
are derived from the new DMS equation.
2.3. Theory and derivation of curve fitting

Nonlinear (multiparameter) regressions of experimental data

have been performed using a common algorithm based on the
NewtoneRaphson method. Two parameters, the sum of squares
due to error (SSE) and R2 are utilized to estimate the goodness of
fit statistics for parametric models [25]. SSE measures the total
deviation of the predicted values from the experiment values,
which is also called the summed square of residuals, i.e.
X
n
 2
SSE ¼ yi  byi ð50Þ
i¼1

where yi, byi are the experiment value and predicted value,
respectively.
R2 is the square of the correlation between the response
values and the predicted response values, which is defined as
the ratio of the sum of squares of the regression (SSR) and
the total sum of squares (SST), i.e. Fig. 10. Water vapor sorption isotherms of nylon 6,6 at 25  C, 35  C, and
45  C respectively. (B) Sorption at 25  C; (6) sorption at 35  C; (7) sorption
SSR at 45  C. The solid lines are derived from the new DMS equation. All the data
R2 ¼ ð51Þ
SST are from Ref. [26].
 H. Feng / Polymer 48 (2007) 2988e3002 2997

Table 1
Estimated values of the parameters and fitting efficiencies of the new DMS
model for water vapor sorption isotherms in nylon 6,6 at 25  C, 35  C, and
45  C
Temperature ( C) Cp (fraction dry basis) k0 A0 R2
25 0.06317 0.6159 1.02 0.9972
35 0.05574 0.6244 1 0.9992
45 0.05209 0.6121 1 0.9991

The estimated values for the new DMS model parameters

are listed in Table 1.
Since the glass transition temperature (Tg) of nylon 6,6 is
45  C, the nylon should be in its glassy state at 25  C and
35  C. However, A0 ¼ 1 for the sorption isotherms at 35  C
and 45  C, suggesting the polymer is in rubbery state for
such cases.
The polymer is indeed in the rubbery state. This is due to
the fact that the sorbed water acted as an effective plasticizer
in depressing the glass transition temperature (Tg) of the poly-
mer and it has great influence on Tg. Loong et al. [26] mea-
sured Tg of nylon 6,6 at various water contents using
differential scanning calorimetry. The results are presented
in Fig. 11. Tg ¼ 25.7  C at a ¼ 0.2. Therefore, the nylon 6,6
film at 35  C and 45  C actually at its rubbery state. A0 is
slightly greater than 1 at 25  C. This may due to measurement
or fitting errors or due to the fact that there are still a few
microvoids in the polymer at 25  C which is much closer to Tg.
This case illustrates that if A0 ¼ 1, the polymer is in its
rubbery state. Furthermore, both Fig. 10 and Table 1 suggest
that the new DMS model describes the experiment data quite
satisfactorily. All of these validate the assumptions of the
new DMS model.
3.1.2. Example 2
Fig. 12 gives sorption isotherms of ethanol vapor in three
polymers (polyurethaneimide (PUI) block copolymers), com-
positions of which are shown in Table 2. Table 3 shows the
Fig. 12. Sorption of ethanol in three polar elastomers at 30  C. (6) Ethanol
sorption in PTMG650eMDIeAP; (>) ethanol sorption in PPG725eMDIe
AP; (7) ethanol sorption in PCL1250eMDIeAP. All of them are the exper-
iment data from Ref. [27]. The solid lines are the results of the sorption
isotherms fitted by the new DMS model.
estimated values of the parameters and fitting efficiencies of
the new DMS model for ethanol vapor sorption in the three
PUI block copolymers. Since the three polymers are rubbery,
parameter A0 should be 1 if the new DMS model is valid. Table
3 clearly shows that all the parameters A0 s of the isotherms are
equal to 1. Besides, the new DMS model fits the experiment
data well. This example illustrates the rationality of the new
DMS model.
3.1.3. Other examples
The same conclusions also can be reached by other
polymer/vapor systems. Table 4 gives estimated values of
the parameters of the new DMS model for five rubbery poly-
mer/vapor systems. Experiment data fitted by the new DMS
model were obtained from Kim et al. [28].
Three of the A0 values exactly equal 1. Two of the A0 values
are slightly greater than 1, which may result from the measure-
ment or fitting errors. Moreover, according to the square of the
correlation, R2, between the experiment data and the predicted
values, shown in Table 4, the model fits extremely well with

Table 3
Estimated values of the parameters and fitting efficiencies of the new DMS
model for ethanol vapor sorption in three PUI block copolymers
Polymer Cp (volume fraction) k0 A0 R2
PCL1250eMDIeAP 0.0162 0.9022 1 0.9957
PPG725eMDIeAP 0.0377 0.8668 1 0.9882
Fig. 11. Glass transition temperature Tg of nylon 6,6 film as related to moisture
PTMG650eMDIeAP 0.1501 0.5249 1 0.9930
content.

Table 2
Chemical compositions of the three PUI copolymers
Abbreviation Soft segment precursors Mn Hard block
PTMG650eMDIeAP Polytetramethyleneglycol 650 Macrodiisocyanate/pyromellitic anhydride (MDI/AP): pyromellitimideurethane
PPG725eMDIeAP Polypropyleneglycol 725
PCL1250eMDIeAP Polycaprolactonediol 1250
2998 H. Feng / Polymer 48 (2007) 2988e3002

Table 4
Estimated values of the parameters and fitting efficiencies of the new DMS model for five polymer/vapor systems
Polymer Vapor Temperature ( C) Cp (weight fraction) k0 A0 R2
Poly(dimethylsiloxane) (PDMS) n-heptane 30 0.2017 0.7203 1.001 0.9999
PDMS n-octane 30 0.1805 0.7484 1 0.9996
PDMS n-nonane 40 0.1934 0.7368 1 0.9955
Poly(propylene oxide) (PPO) Methanol 30 0.0902 0.8660 1.006 0.9996
PPO Ethanol 30 0.1359 0.8333 1 0.9970

the experiment data. These results validate the assumptions of
the new DMS model.
If the conventional DMS model is used for the rubbery
polymer, cH ¼ 0, c ¼ cD ¼ kD0a which is linear. It does not
agree with the experiment data.
3.2. Application of the new DMS model to vapor sorption
in glassy polymers
3.2.1. Application of the new DMS model to Type 1 vapor
sorption isotherms
Water sorption in two polyimide copolymers is taken as an
example of this type. Fig. 13 shows isotherms of water vapor
sorption in the two polyimide copolymers, pyromellitic dia-
nhydride-4,40 -diaminophenylsulfone (50 mol%)/4,40 -oxydia-
niline (50 mol%) (PMDAe50DDS/50ODA), and biphenyl
tetracarboxylic dianhydridee50 mol% DDS/50 mol% ODA
(BPDAe50DDS/50ODA) at 35  C. Data for these systems
were obtained from Huang et al. [2]. These sorption data are
fitted by the new DMS and the CDMS models. Tables 5 and
6 list estimated values of the parameters and the fitting
efficiencies of the two models for water vapor in these copol-
ymers at 30  C.
All the isotherms are concave towards the activity axis over
the entire activity range. Inflection points of the two isotherms
fitted by the new DMS model are at a ¼ 20 and 1.19, respec-
tively, also indicating that all the isotherms are Type 1.
It is well known that this type of vapor sorption data can be
accurately described by CDMS model, which is typically used
to model gas or vapor sorption in glassy polymers. This exam-
ple confirms this result. Furthermore, the isotherms obtained
by the CDMS and the new DMS models almost coincide
with each other, indicating that both of them fit extremely
well with the experiment data, as displayed in Fig. 13. Table
6 also suggests that the new DMS model is as effective as
the CDMS model in fitting Type 1 vapor sorption isotherms.
Therefore, both the CDMS and the new DMS models
describe Type 1 vapor sorption isotherms in glassy polymers
extremely well.
3.2.2. Application of the new DMS model to Type 4 vapor
sorption isotherms
Since this type of isotherms looks like Type 1, researchers
such as Okuno et al. [11] utilized only the CDMS model to fit
the experiment data. Present work fits the new DMS and
CDMS models to the experiment data referenced from Okuno
et al. [11]. Fig. 14 presents water and ethanol vapor sorption
isotherms in PVC at 40  C fitted by the new DMS and
CDMS models. Tables 7 and 8 give the estimated values of
the parameters and the fitting efficiencies of the two models,
respectively.
The inflection points are at a ¼ 0.57 and 0.44, respectively,
for water and ethanol vapor sorption in PVC at 40  C, suggest-
ing that the isotherms are actually Type 2. Both Fig. 14 and

Table 5
Estimated values of the parameters of the new DMS and the CDMS models for
water vapor in dense polyimide copolymers at 30  C
Polyimide New DMS model CDMS model
0 0
k A Cp kD0 CH0 b0
(wt%) (wt%) (wt%)
PMDAe50DDS/50ODA 0.01212 15.85 26.6 0.0009924 30.9 0.1651
BPDAe50DDS/50ODA 0.2091 7.4 2.93 1.473 1.38 2.548

Table 6
Comparison of the new DMS model with the CDMS model in fitting water
Fig. 13. Water vapor sorption isotherms in two polyimide copolymers, vapor in dense polyimide copolymers at 30  C
PMDAe50DDS/50ODA and BPDAe50DDS/50ODA at 30  C. (B) Experi-
ment data for PMDAe50DDS/50ODA; (6) experiment data for BPDAe Copolymer Model SSE R2
50DDS/50ODA. The blue dot line and red dash line represent the best fits PMDAe50DDS/50ODA New DMS 3.5E3 0.9998
of the new DMS model and the CDMS model, respectively. Data for these sys- CDMS 3.5E3 0.9998
tems were obtained from Huang et al. [2]. (For interpretation of the references
BPDAe50DDS/50ODA New DMS 1.1E2 0.9979
to color in this figure legend, the reader is referred to the web version of this
CDMS 9.5E3 0.9981
article.)
 H. Feng / Polymer 48 (2007) 2988e3002 2999

Table 8
Fitting efficiencies of the new DMS and the CDMS models for sorption
isotherms of water and ethanol vapors in PVC at 40  C
Vapor Model SSE R2
Water vapor New DMS model 2.09E9 0.9997
CDMS 6.99E9 0.9989
Ethanol vapor New DMS model 3.86E7 0.9993
CDMS 2.24E6 0.9959

Both Fig. 15 and Table 10 demonstrate that the new DMS

model agrees extremely well with the experiment data for such
cases.
Another example of Type 2 isotherm is methanol vapor
sorption in cellulose acetate at 25  C. Sorption data are fitted
by three models, new DMS, ENSIC and CDMS, as illustrated
Fig. 14. Sorption isotherms of water and ethanol vapors in PVC at 40  C. (B) in Fig. 16. Estimated values of the parameters of the three
Experiment data for water vapor sorption; (6) experiment data for ethanol va-
models for the sorption are listed in Table 11. Their fitting
por sorption; all the data obtained from Okuno et al. [11]. The blue dot line
and red dash line represent the best fits of the new DMS model and the
CDMS model to the experiment data, respectively. The unit of vertical axis
is 103 g/cm3 polymer for water vapor sorption and 102 g/cm3 polymer for
ethanol vapor sorption. Inflection points are at a ¼ 0.57 and 0.44 for water va-
por and ethanol vapor sorption in the polymer, respectively. (For interpretation
of the references to color in this figure legend, the reader is referred to the web
version of this article.)

Table 8 indicate that the new DMS model is superior to the

CDMS model in fitting Type 4 isotherms.

3.2.3. Application of the new DMS model to Type 2 vapor

sorption isotherms
Characteristics of water sorption in three glassy copolymers
are investigated in this section. The three sulfonated polyimide
copolymers were synthesized by copolymerization of 1,4,5,8-
naphthalenetetracarboxylic dianhydride (NTDA) and 2,20 -bis
(4-aminophenoxy)biphenyl-5,50 -disulfonic acid (oBAPBDS)
Fig. 15. Water vapor sorption in the three sulfonated polyimide copolymers at
(NTDA-oBAPBDS), copolymerization of NTDA and 4,40 - 50  C. (B) Experiment data for sorption in NTDAeoBAPBDS; (>) experi-
bis(4-aminophenoxy)biphenyl-3,30 -disulfonic acid (BAPBDS) ment data for sorption in NTDAeBAPBDS; (,) experiment data for sorption
(NTDA-BAPBDS), and copolymerization of NTDA and in NTDAeoBAPBDS/mBAPPS; solid lines represent fit of the new DMS
oBAPBDS/bis[4-(3-aminophenoxy)-phenyl] sulfone (mBAPPS) model to the experiment data which were from Ref. [29].
(NTDA-oBAPBDS/mBAPPS(2/1)), respectively.
Water vapor sorption isotherms in the three sulfonated poly- Table 9
imide copolymers at 50  C fitted by the new DMS model are pre- Estimated values of the parameters of the new DMS model for water vapor
sorption in the three sulfonated polyimide copolymers at 50  C
sented in Fig. 15. The inflection points are at a ¼ 0.31, 0.30,
0.30, respectively, for water sorption in NTDA-oBAPBDS, Copolymer New DMS model
NTDA-BAPBDS and NTDA-oBAPBDS/mBAPPS(2/1) at k0 A0 Cp (g/100 g dry polymer)
50  C, suggesting the isotherms are real Type 2. NTDAeoBAPBDS 0.7912 12.92 13.97
Tables 9 and 10 give the parameters and the fitting efficien- NTDAeBAPBDS 0.7537 22.57 12.15
cies of the new DMS for water vapor sorptions in the three NTDAeoBAPBDS/ 0.8089 14.42 8.865
mBAPPS(2/1)
sulfonated polyimide copolymers at 50  C, respectively.

Table 7
Estimated values of the parameters of the new DMS and the CDMS models for sorption isotherms of water and ethanol vapors in PVC at 40  C
Vapor New DMS model CDMS model
k0 A0 Cp (g/cm3 polymer) kD0 (g/cm3 polymer) CH0 (g/cm3 polymer) b0
Water vapor 0.43 14.14 1.918E3 22.96E3 3.211E3 90.33
Ethanol vapor 0.5633 10.5 12.73E3 2.01E3 1.094E3 12.06
3000 H. Feng / Polymer 48 (2007) 2988e3002

Table 10 Table 12
Goodness-of-fit evaluations for water vapor sorption in the three sulfonated Fitting efficiencies of the three models for methanol vapor sorption in
polyimide copolymers at 50  C using the new DMS model cellulose acetate at 25  C
Copolymer SSE R2 Model SSE R2
NTDAeoBAPBDS 2.11 0.9993 New DMS model 9.94E5 0.9987
NTDAeBAPBDS 2.46 0.9991 CDMS 1.09E3 0.9859
NTDAeoBAPBDS/ 2.96 0.9985 ENSIC 6.92E4 0.9911
mBAPPS(2/1)

Both Fig. 17 and Table 14 clearly illustrated that the new

efficiencies are given in Table 12. The inflection point is at
a ¼ 0.33.
Both Fig. 16 and Table 12 illustrated that the new DMS
model is obviously superior to the other models, CDMS and
ENSIC for this type of sorption.
3.2.4. Application of the new DMS model to Type 5 vapor
sorption isotherms
Water vapor sorption in cellulose acetate at 25  C is inves-
tigated in this section. Since this type of isotherms look like
Type 3, such as water vapor sorption in cellulose acetate at
25  C, as illustrated in Fig. 17, Perrin et al. [10] utilized the
ENSIC model to fit the experiment data. Present work fits
the ENSIC, CDMS and new DMS model to the experiment
data from Perrin et al. [10].
Fig. 17 displays water vapor sorption isotherms in cellulose
acetate at 25  C fitted by three models, CDMS, ENSIC and
new DMS. Tables 13 and 14 give the estimated values of the
parameters and the fitting efficiencies of the three models.
DMS model is obviously superior to the other models,
CDMS and ENSIC, for this type of sorption.
3.2.5. Application of the new DMS model to Type 3 vapor
sorption isotherms
Ethanol vapor sorption in cellulose acetate at 25  C is
investigated in this section. Fig. 18 displays ethanol vapor
sorption isotherms in the cellulose acetate at 25  C fitted by
three models, CDMS, ENSIC and new DMS. Tables 15 and
16 give the estimated values of the parameters and the fitting
efficiencies of the three models.
Although the sorption isotherm of water vapor is an upward
curve, the parameter A0 equals 1.87, indicating that the water
sorption in the microvoids of the polymer still exists and the
polymer is in its glassy state. Cellulose acetate at 25  C is in
fact in its glassy state. This illustrates that the new DMS model
is reasonable.
Both Fig. 18 and Table 16 show that the new DMS model
agrees slightly better than the ENSIC model and obviously
better than the CDMS model for such case.

Fig. 16. Sorption isotherms for methanol vapor in cellulose acetate at 25  C.
(B) Experiment data from Perrin et al. [10]. The blue solid line, red dash
line and green dot line represent the best fit of the three models, new DMS,
ENSIC and CDMS, respectively. (For interpretation of the references to color
in this figure legend, the reader is referred to the web version of this article.)
Fig. 17. Water vapor sorption in cellulose acetate at 25  C. (B) Experiment
data from Perrin et al. [10]. The blue, red and green lines represent the best
fit of the three models, new DMS, ENSIC and CDMS, respectively. The inflec-
tion point is at a ¼ 0.14. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

Table 11
Estimated values of the parameters of the three models, new DMS, CDMS and ENSIC for methanol vapor sorption in cellulose acetate at 25  C
New DMS model CDMS model ENSIC model
0 0
k A Cp (volume fraction) kD0 (volume fraction) CH0 (volume fraction) b0 ks kp
0.7058 5.24 0.1228 0.3128 0.05381 0.01371 0.2557 0.7848
 H. Feng / Polymer 48 (2007) 2988e3002 3001

Table 13
Estimated values of the parameters of the three models, new DMS, CDMS and ENSIC for water vapor sorption in cellulose acetate at 25  C
New DMS model CDMS model ENSIC model
0 0
k A Cp (volume fraction) kD0 (volume fraction) CH0 (volume fraction) b0 ks kp
0.845 2.59 0.03532 0.1238 0.006557 0.003025 0.2557 0.7848

Table 14 Table 16
Fitting efficiencies of the three models for water vapor sorption in cellulose Comparison among the fitting efficiencies of the three models for ethanol
acetate at 25  C vapor sorption in cellulose acetate at 25  C
Model SSE R2 Model SSE R2
New DMS model 1.49E5 0.9994 New DMS model 2.14E5 0.9993
CDMS 3.76E3 0.8469 CDMS 4.58E4 0.9846
ENSIC 2.55E5 0.9896 ENSIC 4.20E5 0.9986

Table 15
Estimated values of the parameters of the three models, new DMS, CDMS and ENSIC for ethanol vapor sorption in cellulose acetate at 25  C
New DMS model CDMS model ENSIC model
0 0
k A Cp (volume fraction) kD0 (volume fraction) CH0 (volume fraction) b0 ks kp
0.4729 1.87 0.2125 0.2225 0.05125 0.01732 0.1719 0.86

Table 17
Comparison of the new DMS model with the CDMS and the ENSIC models
for fitting all types of vapor sorption isotherms in glassy polymers
Model Type 1 Type 2 Type 3 Type 4 Type 5
New DMS Very good Very good Very good Very good Very good
CDMS Very good Poor Poor OK Poor
ENSIC Poor Poor Very good Poor OK

glassy polymers, one of which occurs in the matrix region

k0 a
of the glassy polymers and follows c1 ¼ Cp 1k 0 a which is a

downward curve similar to that in rubbery polymers; the0 other

ðA 1Þk 0 a
happens in the microvoids and follows c2 ¼ Cp 1þðA 0 1Þk 0 a

which is an upward curve. The total sorption amount of vapor

Fig. 18. Ethanol vapor sorption in cellulose acetate at 25  C. (B) Experiment
data from Perrin et al. [10]. The blue dash line, red dot line and green solid line
represent the best fit of the three models, new DMS, ENSIC and CDMS, re-
spectively. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.)
Comparisons of the new DMS model with the CDMS and
the ENSIC models for all types of vapor sorption isotherms
in glassy polymers are summarized in Table 17. The table
clearly shows that the new DMS model fits the experiment
data extremely well for all types of vapor sorption isotherms
in the glassy polymers while the other models can describe
only one type of the isotherms very well.
4. Conclusions
1. Like the CDMS model, the new DMS model for vapor
sorption in glassy polymers assumes that two species of the
sorbed molecules contribute to the penetrant concentration in
molecules in the glassy polymers is equal to the sum of the
two species.
2. The new DMS model is reasonable for vapor sorption in
glassy polymers.
3. The new DMS model fits the experiment data extremely
well for all types of vapor sorption isotherms in the glassy
polymers while the CDMS model can describe only Type 1 iso-
therms very well and the ENSIC model agrees extremely well
with only Type 3 isotherms for they have no inflection points.
References
[1] Serad GE, Freeman BD, Stewart ME, Hill AJ. Polymer 2001;42:
6929e43.
[2] Huang Jingui, Cranford Richard J, Matsuura Takeshi, Roy Christian.
J Appl Polym Sci 2003;87:2306e17.
[3] Alentiev AYu, Shantarovich VP, Merkel TC, Bondar VI, Freeman BD,
Yampolskii YuP. Macromolecules 2002;35:9513e22.
[4] Huang Jingui, Cranford Richard J, Matsuura Takeshi, Roy Christian.
J Membr Sci 2004;241:187e96.
[5] Krykin MA, Zinovieva OM, Zinoviev AB. J Polym Sci Part B Polym
Phys 1999;37:2314e23.
3002 H. Feng / Polymer 48 (2007) 2988e3002
[6] Gocho Hiromi, Tanioka Akihiko, Nakajima Toshinari. J Colloid Interface
Sci 1998;200:155e60.
[7] Watari Tatsuya, Wang Hongyuan, Kuwahara Katsunari, Tanaka Kazu-
hiro, Kita Hidetoshi, Okamoto Kenichi. J Membr Sci 2003;219:137e47.
[8] Kamaruddin HD, Koros WJ. J Polym Sci Part B Polym Phys 2000;38:
2254e67.
[9] Detallante Virginie, Langevin Dominique, Chappef Corinne, Metayer
Michel, Mercierb Regis, PinI Michel. Desalination 2002;148:333e9.
[10] Perrin Laurent, Nguyen Quang Trong, Sacco Daniel, Lochon Pierre.
Polym Int 1996;42:9e16.
[11] Okuno Hiroshi, Renzo Kazuhiko, Uragami Tadashi. J Membr Sci 1995;
103:31e8.
[12] Koros WJ, Paul DR. J Polym Sci Polym Phys Ed 1976;14:687e702.
[13] Chiou JS, Paul DR. J Membr Sci 1987;32:195.
[14] Koros WJ, Paul DR. J Polym Sci Polym Phys Ed 1978;16:1947e63.
[15] Koros WJ, Chan AH, Paul DR. J Membr Sci 1997;2:165e90.
[16] Bondar VI, Kamiya Y, Yampolskii YP. J Polym Sci Part B Polym Phys
1996;34:369e78.
[17] Pellegrino J, Radebaugh R. Macromolecules 1996;29:4985e91.
[18] Barbari TA, Koros WJ, Paul DR. J Polym Sci Polym Phys Ed 1998;26:
729.
[19] Favre E, Clement R, Nguyen QT, Schactzel P, Neel J. J Chem Soc
Faraday Trans 1993;89:4339e53.
[20] Favre E, Nguyen QT, Clement R, Neel J. J Membr Sci 1996;117:227e36.
[21] Timmermann EO. J Chem Soc Faraday Trans 1 1989;85(7):1631e45.
[22] Timmermann EO. Colloids Surf A 2003;220:235e60.
[23] Jonquieres Anne, Fane Anthony. J Appl Polym Sci 1998;67:1415e30.
[24] Timmermann EO, Chirife J, Iglesias HA. J Food Eng 2001;48:19e31.
[25] Curve fitting toolbox user’s guide. The Math Works, Inc.
[26] Loong TL, Britt IJ, Marvin AT. J Appl Polym Sci 1999;71:197e206.
[27] Jonquieres Anne, Perrin Laurent, Arnold Stephanie, Lochon Pierre.
J Membr Sci 1998;150:125e41.
[28] Kim Jisoo, Choi Eun-hyun, Yoo Ki-Pung, Lee Chul Soo. Fluid Phase
Equilib 1999;161:283e93.
[29] Guo XX, Fang JH, Tanaka Kazuhino, Kita Hidetoshi, Okamoto Kenichi.
J Polym Sci Polym Chem Ed 2004;42:1432e40.